Skip to content
View 5ham5h33r's full-sized avatar
7 followers · 6 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Block or report 5ham5h33r

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
5ham5h33r/README.md

Shamsheer

MS Computer Science @ USC (May 2027) · AI Research Intern @ Centific · ex-Data Scientist / ML-AI Engineer @ ION Group

I build production AI systems — and the evaluation infrastructure that tells you whether they actually work. Agentic pipelines, real-time voice, LLM benchmarking, and the data engineering underneath.

Featured Work

GiveMeVoice — a voice agent that places and answers real phone calls

Bridges Twilio Media Streams and the OpenAI Realtime API over 8 kHz μ-law audio, end to end. Delivers a live transcript, in-call translation, and a structured JSON call-outcome summary in the user's own language. Ships with a zero-cost mock mode so the full UI can be rehearsed without Twilio or OpenAI.

▶ Watch a real call — demo video

Node.js · Express · WebSockets · Twilio Voice · OpenAI Realtime

SilhouetteAI — on-device PII redaction before your prompt leaves the browser

A browser extension that intercepts prompts headed for ChatGPT, Claude, Gemini and Copilot and redacts personally identifiable information using a transformer NER model running entirely locally via ONNX Runtime and WASM — prompt contents are never sent anywhere for analysis. Reversible tokenization restores the original values in the rendered DOM.

ONNX Runtime Web · WASM · Transformer NER · Chrome Extensions · IndexedDB

Dysarthria Detection — speech clarity enhancement for dysarthric speakers

A CNN+LSTM model in TensorFlow/Keras detecting dysarthric speech at 91% accuracy, wired into a Whisper speech-to-text and TTS re-synthesis pipeline behind a live Streamlit app.

TensorFlow · Keras · Whisper · Streamlit

BookYourShow — location-aware event discovery on Android and web

Kotlin and Jetpack Compose app over the Ticketmaster Discovery and Google Places APIs, with network calls and UI state orchestrated through a Repository pattern using Retrofit and Coroutines. Favorites sync via MongoDB Atlas to a companion web app on GCP.

Kotlin · Jetpack Compose · Retrofit · Coroutines · MongoDB Atlas · GCP

What I work with

Languages — Python · SQL (Databricks, PostgreSQL, Oracle, MySQL) · Java · Kotlin · JavaScript · C++

AI & ML — Agentic pipelines · LangChain · RAG · LLM evaluation & benchmarking · PyTorch · TensorFlow · PySpark ML · MLflow

Data & Cloud — Databricks · Unity Catalog · Spark · Azure · GCP · Docker · CI/CD

Portfolio · LinkedIn · shamsheerkhalid99@gmail.com

Pinned Loading

  1. GiveMeVoice GiveMeVoice Public

    Voice agent that places and answers real phone calls on your behalf - Twilio Media Streams bridged to OpenAI Realtime over 8 kHz u-law audio, with live transcript, in-call translation, and a struct…

    JavaScript

  2. SilhouetteAI SilhouetteAI Public

    Browser extension that catches PII in your prompt and lets you redact it before it leaves your machine - local detection, reversible tokenization, zero network calls.

    JavaScript 1

  3. 5ham5h33r.github.io 5ham5h33r.github.io Public

    Portfolio page created to display my information and showcase my skills and projects.

    JavaScript

  4. BookYourShow BookYourShow Public

    Location-aware event discovery on Android - Kotlin and Jetpack Compose over the Ticketmaster Discovery and Google Places APIs, with favorites synced via MongoDB Atlas to a companion web app on GCP.

    Kotlin